we’re bringing 20 cracked builders into one villa. whether you’re a coder, artist, tinkerer, policy maker, hardware geek — or just someone chasing weird ideas. if that sounds like you, come build with us. apply now:

RT @localhostIND: nothing like spending time with talented builders, sharing wins and losses. push day update - week 1 .

Join millions who have switched to Grok.

been cooking up something since last couple of months.Q4 is gonna be wild! 👀. designing to techno hits different 🎧

50 days ahead!.

when you work hard and party harder at the same time @localhostIND

RT @abhayshenoyx: @localhostIND infinite money card part 2. delhi ncr chapter here we goo!!

that is my framework. what do you think? would love to hear counter opinions or what bottlenecks i’m missing.​​​​​​​​​​​​​​​​. reading “MACHINES OF LOVING GRACE” by dario amodei helped me a lot in framing this thinking.

taking a top-down approach to understand what bottlenecks remain post-agi helps identify problems actually worth solving. instead of building for today’s models, one can look at tomorrow’s constraints and solve for those.

8. COORDINATION STILL SUCKS. agi can’t make governments functional. can’t fix your homeowners association. human groups failing to coordinate due to personal interests this blocks everything else.

7. OUR BRAINS ARE THE BOTTLENECK. human brain: 20 watts, processes ~50 bits per second.gpt-4: megawatts, processes terabits per second. the interface between human and ai is pathetically slow.

6. TRUTH BECOMES SCARCE. when anyone can fake anything perfectly, what’s real?.deepfakes, generated text, synthetic data everywhere. proving authenticity becomes crucial infrastructure.

5. PHYSICS HAS HARD LIMITS. can’t go faster than light. chips can only pack so many transistors. every computation needs minimum energy. no amount of intelligence breaks thermodynamics.

4. HUMANS BLOCK GOOD SOLUTIONS. nuclear power works great. we barely use it. supersonic flight worked. we banned it. agi might propose perfect solutions that humans regulators will take years to accept.

3. SOME THINGS STAY UNPREDICTABLE. even perfect ai can’t predict chaotic systems much better. weather beyond 10 days. stock market crashes. where a leaf falls. uncertainty is fundamental, not a compute problem.

2. YOUR DATA IS TRAPPED. 99% of world’s data sits in company servers and personal devices (proprietary data). openai doesn’t yet have YOUR emails, YOUR medical records, YOUR code. we need personal llms. company llms. trained on private data.

1. SPEED OF PHYSICAL WORLD. ai will probably design a cancer drug in just days. testing it might still take years. building a chip fab takes 5 years. no ai changes that. experiments need time. hardware needs time. humans need time.

if agi hits by 2027 (plus/minus 1 year), why build something that’ll be obsolete?. been thinking - what problems remain even AFTER we solve intelligence and actions (agents) ?.

i ran a thought experiment, what problem statements should one focus on in the age of ai. every 6 months a new model drops. kills half the wrapper startups.

RT @abhayshenoyx: we're looking for the indian @im_roy_lee .a home for the delusional.

RT @suhasasumukh: mann ki baat.

RT @abhayshenoyx: calling all misfits. if you have an idea or are already building but don’t feel like you’re in the right environment, th….